Voice is ( now, but not before) on the x axis as more compute intense than MCHR.
They have switched places. 


This helped me with the following..


The compute intensity of an algo is NOT a constant.
If you ADD in more stuff like AEC.
those additional algos add to the compute intensity.

Add in one more sensor and things shift...
that intelligent power, where QUIK uses the accelerometer
nice use of another sensor and algo?
And a reason to use a HUB over a microphone that cannot make use
of an accelerometer?


That indoor location is waay to the edge. Its 10 axis fusion and just going from 9 to 10 adds
something like 30% more compute intensity...


So as I track stuff along every time they add to make a device better( using AEC as a mental model)
it is another algo adding compute intensity......we want MORE sensors added to the mix.